A tipical call center consists of telephone trunk lines,Interactive Voice Response,IVR for short ,and Agents.The scheme of this thesis is that we consider a call center by using Markov model.Because the models that we consider are all irreducible finite MC,stationary distribution must exist.In Chapter 2 we consider a queue in condem with two nodes.First,by means of constructing a two-state MC I'll prove that when the impatient time of the customers is infinite ,its stationary probability has a product-form solution.Second,using the same method I'll prove that when the impatient time of the customers has an exponential distribution,its stationary probability also has a product form .Hence, a series of performance indexes in a call center can be deduced, inclusive of busy probability, waiting time distribution,and delay probability.Therefore,the model will be more closer to the actual situation.In Chapter 3 we consider a queue in condem with three nodes. We'll better the model further—consider that customers maybe retrial when they get busy signals .Although the product-form solution of the stationary state probability will be destroyed, we still can give it a QBD algorithm . |